Multi-task feature selection

By Guillaume Obozinski, B. Taskar, Michael I. Jordan. 2006

Subfields: Representation & Self-Supervised Learning

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Representation & Self-Supervised Learning

1. Qiang Yang
2. Yu Zhang
3. Yu Zhang
4. Jieping Ye
5. Dit-Yan Yeung
6. Dacheng Tao
7. Jiayu Zhou
8. Mingsheng Long

Citations in the corpus

• An Overview of Multi-task Learning (2018, methodology): “Concrete instances of the (cid:2) p , q regularization include the (cid:2) 2,1 regularization proposed in [8,9] and the (cid:2) ∞ ,1 regularization proposed in [10].”
• Cross-Stitch Networks for Multi-task Learning (2016, background): “The term multi-task learning (MTL) itself has been broadly used [2, 14, 28, 42, 54, 55] as an umbrella term to include representation learning and selection [4, 13, 31, 37], transfer learning [39, 41, 56] etc.”
• A Survey on Multi-Task Learning (2017, methodology): “[24], [25] are among the first to study the multi-task feature selection (MTFS) problem based on the `2,1 norm with the objective function formulated as”
• Multi-Stage Multi-Task Feature Learning (2012, neutral): “It has been successfully applied to many applications such as speech classification (Parameswaran and Weinberger, 2010), handwritten character recognition (Obozinski et al., 2006; Quadrianto et al., 2010) and medical diagnosis (Bi et al., 2008).”
• Nonparametric Bayesian feature selection for multi-task learning (2011, unknown): “MODEL FORMULATION We consider M tasks.”
• On Multiplicative Multitask Feature Learning (2014, methodology): “These methods employ a regularizer based on the so-called (cid:96) 1 ,p matrix norm [12, 13, 15, 22, 24] that is the sum of the (cid:96) p norms of the rows in a matrix.”
• Transfer learning for image classification with sparse prototype representations (2008, methodology): “Existing methods for transfer learning often learn a prior model or linear manifold over classifier parameters [8, 2], discover a sparse set of common features [13, 3, 20, 6], or use a representation based on classifier outputs from related tasks [24], but do not generally take advantage of unlabele”
• Multi-output learning via spectral filtering (2012, methodology): “We compare our approach with a state-of-the-art sparsity-enforcing method (Obozinski et al. 2007), which, being a linear model, requires particular basis fields dictionaries for dealing with non-linear vector fields.”
• Block Variable Selection in Multivariate Regression and High-dimensional Causal Inference (2010, methodology): “When applied to time-evolving social media content, our models yield a new family of causality-based influence measures that may be seen as an alternative to the classic PageRank algorithm traditionally applied to hyperlink graphs.”
• Multi-Modal Multi-Task Learning for Joint Prediction of Multiple Regression and Classification Variables in Alzheimer’s Disease (2011, methodology): “Following the method proposed in (Obozinski et al., 2006), linear models are used to model the multi-task feature selection (MTFS) as below: where is the weight vector for the j -th task on the m -th modality, and x ( m ) is the m -th modal data of a certain subject.”
• Semi-supervised Robust Dictionary Learning via Efficient l-Norms Minimization (2013, methodology): “To address the both issues, we propose to enforce structured sparsity on A using l2,1-norm regularization [12, 2].”
• Simultaneous learning of several materials properties from incomplete databases with multi-task SISSO (2019, background): “32,34,39,40 In order to clarify how the MTL concept can be applied in materials science, let us introduce the showcase examples that will be addressed in the following sections.”
• Automatic Feature Decomposition for Single View Co-training (2011, background): “A similar problem arises in the context of feature selection in multi-task settings (Argyriou et al.; Obozinski et al., 2006), when similar parameter sparsity patterns across different tasks needs to be imposed.”
• Feature Selection Based on Structured Sparsity: A Comprehensive Study (2017, background): “When r = 2, p = 1, it is the l2,1-norm of a matrix, which was first introduced in [67] as rotational invariant l1-norm and also used for multitask learning [68], [69] and semisupervised learning [70].”